地理統計應用於雨量站網設計之研究

Abstract

雨量資料是水資源開發調配、各項用水供需計畫研擬及水利工程規劃設計最重要之依據，為了充分的了解及掌握降雨事件在時間與空間上之變化特性，因此對於雨量站網應做適當之規劃。由於台灣地區的降雨受到氣候及地理環境的影響甚大，造成降雨在時間與空間上分布不均勻的現象，為了考慮到氣候對降雨的影響，因此本研究將降雨型態劃分為年雨量與梅雨、對流、颱風、鋒面之時雨量及季節性雨量等各種類型；又由於降雨量會隨著高程而有所變化，因此本研究考慮到高程對降雨量之影響並移除降雨量隨高程變化之趨勢，更可以清楚的描述出雨量在空間中的變異特性。 在本研究中利用地理統計法評估淡水河流域下游內27個雨量站之分布情形，首先計算每個網格點上之推估誤差落在 範圍內之機率值α且繪製α等值圖，其中 會隨著不同水資源規劃之站網評估標準而有所變動，在進行站網評估時，則利用研究區域內之機率值α大於門檻值之面積百分比，作為停止加站或減站之標準，在本研究中將門檻值定為0.8及0.9，此法可改善前人的相對加減站策略。本研究亦提出一個新的站網分布位置，可適當的觀測各類型降雨，且較原始站網分布型態有較好的空間推估效果。當門檻值為0.8時，梅雨時雨量有8個測站是可優先被移除且利用剩餘之19個基本站可使控制面積百分比達0.5379，在改變測站位置後27個測站之控制面積百分比為0.8791，在各類型降雨中可發現梅雨時雨量具有較大的空間變異性，因此選擇梅雨時雨量之站網分布作為研究區域中之最佳站網位置。

Rainfall data is one of the essential data in hydrological analysis and engineering design including water budget analysis, frequency analysis and stormwater drainage design. Direct measurement of rainfall can only be achieved by raingauges and raingauge networks are often installed to provide temporal and spatial variations of rainfall. However, even though raingauges are capable of measuring rainfall rate in real-time and at very fine resolution in time, the spatial variation of rainfall is still difficult to be characterized without a raingauge network of enough density in space. In addition, selection of raingauge locations is affected by many factors such as accessibility, easiness of maintenance, topographical aspects, etc. Furthermore, the density of a raingauge network is dependent on the time resolution (or scale) of the desired rainfall measurements. For example, for the purpose of water resources planning, observation of monthly or annual rainfall is desired; however, for flood mitigation and forecasting, hourly rainfall must be measured. Hourly rainfall exhibits higher spatial variability and thus, as compared to monthly or annual rainfall, a network of higher density is needed. Therefore, a methodology for raingauge network design and performance evaluation of an existing raingauge network is important in that it can help to understand its capability and the quality of the data it provides. In this study, a geostatistical approach for raingauge network design and evaluation was proposed and a network of 27 raingauges in lower Danshuei River watershed was chosen for evaluation. It first defines the “acceptable precision” for rainfall estimation at ungauged sites using rainfall measurements of the existing raingauges and the ordinary kriging. The rainfall estimates at ungauged sites are considered acceptable if the probability αA that the estimate falls within one standard deviation of the true value exceeds a specified level α, say 0.8 or 0.9. By adopting such a criterion, both the ungauged locations and the area percentage pA within the study area satisfying this criterion can be obtained. A sequential algorithm is also developed to find raingauges that provide only redundant information and can be eliminated or moved. Relocation of such raingauges can also be suggested using a contour map of αA and the area percentage of pA corresponding to the altered network. It was found that, among the four major storm types in Taiwan, hourly rainfall of Mei-Yu exhibited the highest spatial variability and therefore raingauge network evaluation was conducted based on hourly rainfall of Mei-Yu. It was concluded that at α = 0.8, 8 existing raingauges are redundant and only 19 raingauges are needed to achieve pA=0.5379. After relocating the 27 raingauges, the altered network can achieve pA=0.8791.